CUMULATIVE SEMANTIC INHIBITION
Table 3: Comparison of estimated parameters in various linear regression models of the log-
transformed naming latency in the restricted dataset (1056 observations from 24 participants naming
50 items in 10 categories).
Step |
N-model |
trial |
Position of. . . |
Fixed effects co-cat. |
Ordinal within. . . | ||
1st |
1 |
β p |
5.88 ∙ 10-4 2.48 .013 |
4.28 ∙ 10-2 4.13 <.001 | |||
2nd |
2 |
β |
-1.22 ∙ 10-4 -.05 |
2.28 ∙ 10-4 |
6.47 ∙ 10-2 2.56 <.01 |
4.73 ∙ 10-2 3.07 | |
3 |
β p |
7.00 ∙ 10-2 3.56 <.001 |
4.68 ∙ 10-2 4.49 <.001 | ||||
3rd |
4 |
β p |
7.68 ∙ 10-5 |
3.07 ∙ 10-2 2.61 <.01 |
1.38 ∙ 10-2 2.54 | ||
5 |
β |
3.02 ∙ 10-2 2.61 <.001 |
1.47 ∙ 10-2 3.55 <.001 |
a cat. = category
tercept across categories (χ2 (1) = 5.99, p = .014) and a variable ordinal position effect
(χ2(1) = 5.08, p = .025). The estimate of N-model 1 for the Ordinal position effect is
plotted in the leftmost panel of Figure 2.
In contrast to the full dataset analyses, trial position produced a significant effect in
the restricted dataset of N-model 1. This result will be discussed below in relation to the
manipulations concerning the relative position of co-categories. For the rest, the restricted
dataset provides a robust and safe testing ground for the hypotheses stated above about
the mutual influence of co-categories on the cumulative inhibition effect.
Second step: Is there a dependency between co-categories?
The pairs of co-categories summarized in Table 2 may, or may not, be sub-served by
common representations. If two co-categories are independent, naming the items of one of
them should be independent of whether the members of the co-category have been named
before or not. By contrast, if two co-categories have a common underlying representation,
then the category that is presented second should be slowed down by the previous naming
of its co-category. We test this prediction by differentiating co-categories on the basis of
the order in which they were named in the experiment. For this test to be possible, two
precautions need to be taken.
First, the order in which categories were presented across participants has to be